Spindle-Savvy AI Monitoring System Reduces Machining Downtime
On a machining center, the continuous process of high-speed spindle rotation that transfers torque and power to cutting tools eventually takes its toll on the system’s shaft and bearings. This fatigue wear leads to out-of-spec operation and eventual failure. Spindles drift out of alignment and balance over time. They can suffer damage from crashes that bend their shafts, along with increased heat from worn and roughened bearing races. Without the ability to analyze spindle behavior to predict approaching failures, shops can find themselves in the middle of large deadline-sensitive jobs with out-of-commission equipment and needed replacement parts that could take weeks to get.
Spindle downtime costs money in more than one way. The loss of a single machine can put an entire cell out of work or disable a production process that relies on multiple pieces of equipment. Along with the inability to produce parts, meet project deadlines and move on to other work, downtime also can harm supplier/customer relationships. A pattern of equipment failures and missed commitments often leaves a customer looking for alternative suppliers. Shops need a maintenance program that can anticipate problems. New artificial intelligence (AI) technologies can help uncover avoidable problems, making formerly hidden types of damage easy to spot through digital monitoring.
For years, preventive maintenance has relied on a combination of monitoring finished work for discernable flaws and tracking them back to their sources, listening for unusual operating noises and identifying faulty equipment through manual equipment evaluations with testing devices. For instance, ballbar testing identifies positioning problems that appear only when a machining center operates, and drawbar force gages detect reduced pulling force on toolholders, representing gradual decline in spindle performance. These tools require constant diligence to assure that shop personnel use them correctly and consistently, and report problems for prompt attention.
Instead of hoping to avoid failures, shops increasingly look to automated systems and artificial intelligence (AI) technologies for ways to monitor their equipment to report wear before it causes critical faults. Additionally, organized, strategic condition monitoring provides opportunities to plan proactive maintenance so it fits around peak production periods. An AI-based monitoring solution can use a baseline assessment of a spindle to determine when performance is heading toward a critically impaired state. This can reduce downtime and the production of out of tolerance parts.
On the road to automated spindle health assessment, specific types of performance measurement data can reveal impending failures. For example, vibration analysis finds abnormal patterns that correlate with bearing failures. Unless the monitoring system can distinguish between those vibration signatures that spell spindle damage and those that do not, however, it can report false positive conditions that lead to unnecessary repairs.
Third-party monitoring systems include sensors that attach to specific positions on the spindle and wire up to data collection systems that either store data in the shop itself or report it to a remote monitoring site through an Internet connection. Unless these sensors, the software that accesses them and the collection system integrate seamlessly with the machining center, the monitoring system may be unable to detect a useful range of potential problems. Ideally, the machine tool OEM itself creates and installs the health-monitoring system as a seamless component of the equipment. This system would include integrated sensors and a user interface that displays on the machine CNC, with remote monitoring options via wireless access.
Mazak leads in formulating an AI-based system that provides these options on machines equipped with the Smooth CNC. Developed in conjunction with the Industrial AI (IAI) Center at the University of Cincinnati, a leader in predictive analytics, the Mazak Spindle Health Monitoring System uses EDGE Computing and Data Analytics algorithms to model each machine's spindle signature during a 60-second fixed cycle test run, establishing a baseline value for comparison in subsequent tests. Operators can conduct fixed cycle tests at any time and view spindle health on Health Assessment and Fault Diagnosis screens displayed on a machine's Smooth CNC. Predictive analytics diagnose pre-failure conditions before downtime occurs.
In its current configuration, the Mazak Spindle Health Monitoring System uses two sensors that feed high-definition data to an acquisition unit connected to an industrial computer for processing. A Neural Network Self-Organizing Map (SOM), also known as artificial intelligence, compiles a growing profile of each machine, learning to assess its health through features extracted from a growing data set.
As technology continues to revolutionize machining, predictive analytics brings machine learning to spindle-health diagnosis, providing shops with advance warning of conditions that can derail their productivity and profitability. The Mazak Spindle Health Monitoring System represents ongoing development in OEM integration of smart features into high-tech machines. As its capabilities continue to develop toward a production-ready option, it shows that spindle-health guesswork has become a thing of the past.
12/12/2022Enquire about this StoryReturn to News Overviews